Computer vision-based control of an autonomous blimp

The objective of this study is twofold: to approximate a model of a blimp, and to use this model to develop a setup to track a target with the blimp that is outfitted with a wireless camera and radio-controlled propellers. This article presents a powerful method to track any moving or stationary target with an unmanned aerial vehicle by combining the advantages of the proportional derivative (PD) controller, continuously adaptive mean shift (Camshift) algorithm, and pulse width modulation (PWM) method. As a result, it is demonstrated that a decent approximation of blimp behavior is sufficient when using a mathematical model that contains saturation in velocity and actuation. Additionally, a code for the proposed algorithm is developed to capture every frame sample with a frame grabber as a sensor in real time. Once the previously chosen object is tracked, the coordinate data (location information) are transferred to the controller to apply required pulses to DC motors on the blimp. In this paper, the proposed controller is outlined in two steps. Initially, one calculates a PD controller that fulfills the specifications of the mathematical model without saturation. Secondly, PWM is utilized to address the impact of nonlinearities.